Grass: Crowdsourced Data for AI

Introduction

Grass is a decentralized web scraping and data provisioning protocol built on Solana. The network aims to harness unused residential internet bandwidth to aggregate real-time data for large enterprises, AI labs, investment funds, and other data-driven organizations. Grass’s vision is to democratize access to high value web resources by incentivizing a decentralized network of node operators to transparently aggregate web data.

In traditional, centralized web-scraping frameworks, large corporations typically exploit user devices—often discreetly—through terms and conditions in free applications, remote device firmware, or VPN software. Users rarely have insight into these arrangements, and even more sparingly receive financial compensation. In turn, the broader market for web-scraped data, currently dominated by two or three key incumbents, operates under opaque and ethically questionable practices. By contrast, Grass seeks to disrupt this market by realigning incentives such that verified institutions can purchase resources directly from individuals, giving them rewards commensurate with their contribution.

With the extraordinary surge in demand for real-time data from AI models, Grass also serves a second and potentially larger market: live context retrieval (LCR). LCR entails connecting a LLM to current web content so that the model can ingest or reference dynamic data sets to generate accurate and time-sensitive outputs. Grass’s architecture is intended to be a neutral, user-owned network for obtaining real-time web data and verifying data provenance, making it a potentially critical layer for AI systems.

Utilizing publicly available AI data licensing deals, we have estimated Grass’s forward-looking annualized revenue at $26.6M, which could grow as high as $250M by 2029 if the network executes on its scalability goals and the AI training dataset market grows at a projected 27.7% CAGR.

The Market Opportunity

AI has Changed the Data Landscape